Trained a new model on customer feedback data. Accuracy went from 78% to 94% just by cleaning the input data. No architecture changes needed. Garbage in, garbage out is the most underrated lesson in ML.
EW
Log in to reply